Yeah good point. With large enough pixels and pathological color choices you could almost certainly derive the opposite result. It would be interesting to tes

Yeah good point. With large enough pixels and pathological color choices you could almost certainly derive the opposite result. It would be interesting to test it over a remote desktop session where the screen on the device under test is off. That would eliminate a lot of factors related to the display. Presumably you'd see that the network traffic is either larger to begin with, or doesn't compress quite as well, giving you another reason to say "yes, but what if..."